Customer Churn Prediction
No Thumbnail Available
URL
Journal Title
Journal ISSN
Volume Title
Perustieteiden korkeakoulu |
Master's thesis
Authors
Date
2023-10-09
Department
Major/Subject
Data Science
Mcode
SCI3115
Degree programme
Master's Programme in ICT Innovation
Language
en
Pages
61+9
Series
Abstract
The rule of thumb for most businesses has been that the cost of acquiring a new customer is up to five times higher than retaining an existing one. To deal with this, scientists have recurrently tried to predict the possibility of their company's customers churning out based on historical data and predictive modeling techniques. This paper uses data from a Software-as-a-Service company that focuses on streamlining the management of golf clubs and their tournaments. The search aims to find behavioral patterns in terms of product usage belonging to customers who are at risk of terminating their subscriptions. The approach to this task was to first establish the fundamental steps necessary towards the goal, and then branch them out under well-known data science practices. The initial assignment was to efficiently gather all of the relevant information from the company's database, taking into consideration all of its caveats. The next step was set to analyze the intrinsic patterns between the distinct features present in the dataset and extract insights about the customers' behavior and the tendencies of churning out. A better comprehension of the features also allowed for further engineering the dataset, with the goal of having the version that best expresses a customer's feature usage for prediction. The final step was to assess which prediction model between the Random Forest Regressor, the Multi-Layer Perceptron, the Support Vector Regression, and the Fuzzy Linear Regression would be best fitted to the problem. In reality, the three areas were not atomic, and their development was heavily reliant on each other, creating an iterative process that helped answer the research questions. The purpose of the thesis is to allow the company to predict a customer's probability of churn. This way, the organization will be able to make data-driven decisions about where to allocate its resources in order to mitigate possible losses.Description
Supervisor
Manner, JukkaThesis advisor
Ardelean, AlexandruKeywords
churn, user behavior, data engineering, prediction model